Sensitive answers (Free text analysis)
What it is
Sensitive answers is a page in VouchedFor Elevation that uses AI to highlight free text review answers that may indicate a potential complaint or a potentially vulnerable client. It helps you identify answers that may need closer review.
Where to find it: Go to Client actions > Sensitive answers. The page title is Free text analysis.
Who it’s for: Advisers and firms using Elevation who want an early indication of potential complaints or client vulnerability in their review feedback.
What it doesn’t do: The tool does not decide whether a client is vulnerable or whether an answer is a complaint. It highlights answers for you or someone at your firm to read and assess.
The tool cannot guarantee 100% accuracy or coverage. You should still read each review independently, including reviews that have not been flagged.
What gets flagged
Each answer can receive one flag, both flags or neither. Flags are based on explicit wording in the client’s answer; tone alone is not enough.
Potential complaint
Potential complaint flags are aligned to the four FCA Consumer Duty outcomes. The examples below illustrate the types of language that may be flagged.
Area | Example of the kind of language flagged |
Price and value | Fees felt too high, or the client didn't see the value |
Products and services | The product or service didn't do what it should have |
Consumer understanding | Things weren't explained properly or were confusing |
Consumer support | Hard to get through, slow responses, felt ignored |
A passing comment such as "I'd always like to pay less" isn't flagged when the client also says the fee was fair.
Potentially vulnerable
Potential vulnerability flags are aligned to the four drivers of vulnerability in FCA guidance FG21/1:
Driver | Examples |
Health | Physical or mental health conditions, disability, cognitive impairment, care needs |
Life events | Bereavement, divorce, redundancy, becoming a carer, Power of Attorney or probate, where the client describes how it’s affecting them (e.g. struggling to cope, feeling overwhelmed, unsure of next steps) |
Resilience | Low ability to cope with financial or emotional shocks, stress |
Capability | Difficulty with financial matters, low confidence, low literacy or numeracy |
A life event on its own isn’t flagged. For example, “I wanted advice on my retirement” or “I came to you after my divorce” wouldn’t be flagged. “Since my husband died I’ve found it hard to deal with the paperwork” would be, because the client describes the impact on them.
How it works
A client leaves a review with a free text answer.
The answer is scanned by the AI model as it comes in.
If the answer contains a potential complaint or vulnerability indicator, it appears on the Sensitive answers page.
You or someone at your firm reads the answer and decides what action, if any, to take under your firm’s complaints and vulnerable customer processes.
Older reviews are not scanned. Only answers submitted since the feature went live are scanned. Older reviews will not appear on the page. If you would like older reviews to be included, contact us to share your request. A limited look-back may be considered in future, but is not currently available.
Human review is required. Flags are prompts for review, not decisions. Do not base any outcome for a client on the flag alone.
Using the page
The results table shows Adviser, Client, Client answer, Flag and Date. Adviser, Client, Client answer and Date can be sorted; results default to newest first.
If the page shows “No results”, there are no flagged answers for your selected filters and date range. This can happen if your firm is new to Elevation or no answers submitted since launch contain an indicator. Older reviews are not scanned.
Data, privacy and opting out
What’s sent to the AI model: Only the review ID and the answer text are sent. Client names, email addresses, adviser names, firm names and other account details are not sent as separate data fields.
Not used for training: Your firm’s and clients’ data is not used to train the AI model.
Opting out: Contact us if you would like AI features turned off for your firm. We will disable them so your firm’s data is no longer sent to the AI model. This applies to all AI features, not just Sensitive answers.
The Client column displays information from your firm’s own Elevation data. This column is not sent to the AI model.
FAQ
Why can’t I see any results? Only answers submitted since launch are scanned. If none contain an indicator, the page will be empty. Check your date range and adviser or service filters too.
Why are older reviews missing? The feature does not scan reviews submitted before launch. A limited look-back may be considered in future, but is not currently available. Contact us if you would like to request this.
Does a flag mean the client is vulnerable or has made a complaint? No. It means the answer contains language that may need closer review. You or someone at your firm should assess the answer using your firm’s complaints and vulnerable customer processes.
Will it catch everything? No. The tool cannot guarantee complete accuracy or coverage. Continue to read reviews independently and do not rely on this page as your only check.
Why was this answer flagged when it looks fine? The AI highlights possible indicators for review, so some flagged answers will not need action. If a flag looks incorrect, contact us with the review ID so we can investigate.
Why wasn’t this answer flagged? Flags are based on explicit wording in the answer. Indirect hints, tone alone or comments about a previous adviser used only as a comparison are not flagged. If you think an answer has been missed, contact us with the review ID.
Can a single answer have both flags? Yes. An answer with both flags appears once for each flag.
Can we see the prompt or criteria the AI uses? We do not publish the AI prompt. The “What gets flagged” section explains the criteria and gives examples of the types of language that may be highlighted.
Is our clients’ data shared with the AI model? Only the review ID and the answer text are sent. Names and account details are not sent as separate data fields, and the data is not used to train the model.
Can we turn it off? Yes. Contact us and we will turn off all AI features for your firm. It is not currently possible to turn off only Sensitive answers.
Can we choose to have only complaints, or only vulnerability, flagged? Not at the moment. Both types are assessed. Use the filter buttons on the page to view one type of flag.
Need help?
If you have a question about Sensitive answers, want to report an answer that’s been flagged incorrectly or missed, or would like AI features turned off for your firm, email customer_service@vouchedfor.co.uk. If your question is about a specific answer, include the review ID.